---Importing Necessary Files---

In [1]:
import numpy as np
import pandas as pd
import warnings
warnings.filterwarnings('ignore')
from scipy.stats import f_oneway

---Making a DataFrame for the Dataset---

In [2]:
ukroadaccident = pd.read_csv('datasets\\uk_road_accident.csv')

---Checking if the DataFrame is Working---

In [3]:
ukroadaccident
Out[3]:
Index Accident_Severity Accident Date Latitude Light_Conditions District Area Longitude Number_of_Casualties Number_of_Vehicles Road_Surface_Conditions Road_Type Urban_or_Rural_Area Weather_Conditions Vehicle_Type
0 200701BS64157 Serious 5/6/2019 51.506187 Darkness - lights lit Kensington and Chelsea -0.209082 1 2 Dry Single carriageway Urban Fine no high winds Car
1 200701BS65737 Serious 2/7/2019 51.495029 Daylight Kensington and Chelsea -0.173647 1 2 Wet or damp Single carriageway Urban Raining no high winds Car
2 200701BS66127 Serious 26-08-2019 51.517715 Darkness - lighting unknown Kensington and Chelsea -0.210215 1 3 Dry NaN Urban NaN Taxi/Private hire car
3 200701BS66128 Serious 16-08-2019 51.495478 Daylight Kensington and Chelsea -0.202731 1 4 Dry Single carriageway Urban Fine no high winds Bus or coach (17 or more pass seats)
4 200701BS66837 Slight 3/9/2019 51.488576 Darkness - lights lit Kensington and Chelsea -0.192487 1 2 Dry NaN Urban NaN Other vehicle
... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
660674 201091NM01760 Slight 18-02-2022 57.374005 Daylight Highland -3.467828 2 1 Dry Single carriageway Rural Fine no high winds Car
660675 201091NM01881 Slight 21-02-2022 57.232273 Darkness - no lighting Highland -3.809281 1 1 Frost or ice Single carriageway Rural Fine no high winds Car
660676 201091NM01935 Slight 23-02-2022 57.585044 Daylight Highland -3.862727 1 3 Frost or ice Single carriageway Rural Fine no high winds Car
660677 201091NM01964 Serious 23-02-2022 57.214898 Darkness - no lighting Highland -3.823997 1 2 Wet or damp Single carriageway Rural Fine no high winds Motorcycle over 500cc
660678 201091NM02142 Serious 28-02-2022 57.575210 Daylight Highland -3.895673 1 1 Wet or damp Dual carriageway Rural Snowing no high winds Car

660679 rows × 14 columns

---Cleaning the Inconsistencies on the Accident date---

In [4]:
ukroadaccident['Accident Date'] = ukroadaccident['Accident Date'].astype('str')
ukroadaccident['Accident Date'] = ukroadaccident['Accident Date'].str.strip()
ukroadaccident['Accident Date'] = ukroadaccident['Accident Date'].str.replace('/', '-')

---Coverting the Data Type of the Accident Date---

In [5]:
ukroadaccident['Accident Date'] = pd.to_datetime(ukroadaccident['Accident Date'], dayfirst=True, errors='coerce')

---Extracting New Columns from Accident Date---

In [6]:
ukroadaccident['Year'] = ukroadaccident['Accident Date'].dt.year
ukroadaccident['Month_Number'] = ukroadaccident['Accident Date'].dt.month
ukroadaccident['Month'] = ukroadaccident['Accident Date'].dt.month_name()
ukroadaccident['Day'] = ukroadaccident['Accident Date'].dt.day
ukroadaccident['DayofWeek'] = ukroadaccident['Accident Date'].dt.dayofweek  # Monday=0, Sunday=6

---Checking for Null Values---

In [7]:
ukroadaccident.isnull().sum()
Out[7]:
Index                          0
Accident_Severity              0
Accident Date                  0
Latitude                      25
Light_Conditions               0
District Area                  0
Longitude                     26
Number_of_Casualties           0
Number_of_Vehicles             0
Road_Surface_Conditions      726
Road_Type                   4520
Urban_or_Rural_Area           15
Weather_Conditions         14128
Vehicle_Type                   0
Year                           0
Month_Number                   0
Month                          0
Day                            0
DayofWeek                      0
dtype: int64

---Fixing the Null Values---

-Numerical Null Values-

In [8]:
ukroadaccident['Latitude'].mean()
Out[8]:
np.float64(52.553865761110956)
In [9]:
ukroadaccident['Latitude'] = ukroadaccident['Latitude'].fillna(ukroadaccident['Latitude'].mean())

---------------------------------------------

In [10]:
ukroadaccident['Longitude'].mean()
Out[10]:
np.float64(-1.431210368502073)
In [11]:
ukroadaccident['Longitude'] = ukroadaccident['Longitude'].fillna(ukroadaccident['Longitude'].mean())

-Categorical Null Values-

In [12]:
ukroadaccident['Road_Surface_Conditions'] = ukroadaccident['Road_Surface_Conditions'].fillna('unaccounted')

---------------------------------------------

In [13]:
ukroadaccident['Road_Type'] = ukroadaccident['Road_Type'].fillna('unaccounted')

---------------------------------------------

In [14]:
ukroadaccident['Urban_or_Rural_Area'].mode()
Out[14]:
0    Urban
Name: Urban_or_Rural_Area, dtype: object
In [15]:
ukroadaccident['Urban_or_Rural_Area'] = ukroadaccident['Urban_or_Rural_Area'].fillna(ukroadaccident['Urban_or_Rural_Area'].mode()[0])

---------------------------------------------

In [16]:
ukroadaccident['Weather_Conditions'] = ukroadaccident['Weather_Conditions'].fillna('unaccounted')

---Checking if there are still Null Values---

In [17]:
ukroadaccident.isnull().sum()
Out[17]:
Index                      0
Accident_Severity          0
Accident Date              0
Latitude                   0
Light_Conditions           0
District Area              0
Longitude                  0
Number_of_Casualties       0
Number_of_Vehicles         0
Road_Surface_Conditions    0
Road_Type                  0
Urban_or_Rural_Area        0
Weather_Conditions         0
Vehicle_Type               0
Year                       0
Month_Number               0
Month                      0
Day                        0
DayofWeek                  0
dtype: int64

---Checking the Data Type---

In [18]:
ukroadaccident.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 660679 entries, 0 to 660678
Data columns (total 19 columns):
 #   Column                   Non-Null Count   Dtype         
---  ------                   --------------   -----         
 0   Index                    660679 non-null  object        
 1   Accident_Severity        660679 non-null  object        
 2   Accident Date            660679 non-null  datetime64[ns]
 3   Latitude                 660679 non-null  float64       
 4   Light_Conditions         660679 non-null  object        
 5   District Area            660679 non-null  object        
 6   Longitude                660679 non-null  float64       
 7   Number_of_Casualties     660679 non-null  int64         
 8   Number_of_Vehicles       660679 non-null  int64         
 9   Road_Surface_Conditions  660679 non-null  object        
 10  Road_Type                660679 non-null  object        
 11  Urban_or_Rural_Area      660679 non-null  object        
 12  Weather_Conditions       660679 non-null  object        
 13  Vehicle_Type             660679 non-null  object        
 14  Year                     660679 non-null  int32         
 15  Month_Number             660679 non-null  int32         
 16  Month                    660679 non-null  object        
 17  Day                      660679 non-null  int32         
 18  DayofWeek                660679 non-null  int32         
dtypes: datetime64[ns](1), float64(2), int32(4), int64(2), object(10)
memory usage: 85.7+ MB

---Fixing the Data Type---

In [19]:
ukroadaccident['Index'] = ukroadaccident['Index'].astype('category') 
ukroadaccident['Accident_Severity'] = ukroadaccident['Accident_Severity'].astype('category')
ukroadaccident['Light_Conditions'] = ukroadaccident['Light_Conditions'].astype('category')
ukroadaccident['District Area'] = ukroadaccident['District Area'].astype('category')
ukroadaccident['Road_Surface_Conditions'] = ukroadaccident['Road_Surface_Conditions'].astype('category') 
ukroadaccident['Road_Type'] = ukroadaccident['Road_Type'].astype('category') 
ukroadaccident['Urban_or_Rural_Area'] = ukroadaccident['Urban_or_Rural_Area'].astype('category') 
ukroadaccident['Weather_Conditions'] = ukroadaccident['Weather_Conditions'].astype('category') 
ukroadaccident['Vehicle_Type'] = ukroadaccident['Vehicle_Type'].astype('category')
ukroadaccident['Month'] = ukroadaccident['Month'].astype('category')
ukroadaccident['Year'] = ukroadaccident['Year'].astype('category')

---Checking again---

In [20]:
ukroadaccident.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 660679 entries, 0 to 660678
Data columns (total 19 columns):
 #   Column                   Non-Null Count   Dtype         
---  ------                   --------------   -----         
 0   Index                    660679 non-null  category      
 1   Accident_Severity        660679 non-null  category      
 2   Accident Date            660679 non-null  datetime64[ns]
 3   Latitude                 660679 non-null  float64       
 4   Light_Conditions         660679 non-null  category      
 5   District Area            660679 non-null  category      
 6   Longitude                660679 non-null  float64       
 7   Number_of_Casualties     660679 non-null  int64         
 8   Number_of_Vehicles       660679 non-null  int64         
 9   Road_Surface_Conditions  660679 non-null  category      
 10  Road_Type                660679 non-null  category      
 11  Urban_or_Rural_Area      660679 non-null  category      
 12  Weather_Conditions       660679 non-null  category      
 13  Vehicle_Type             660679 non-null  category      
 14  Year                     660679 non-null  category      
 15  Month_Number             660679 non-null  int32         
 16  Month                    660679 non-null  category      
 17  Day                      660679 non-null  int32         
 18  DayofWeek                660679 non-null  int32         
dtypes: category(11), datetime64[ns](1), float64(2), int32(3), int64(2)
memory usage: 61.6 MB

---20 Questions and Insights---

~EDA~

1. What is the most usual vehicle type?

In [21]:
ukroadaccident['Vehicle_Type'].mode()
Out[21]:
0    Car
Name: Vehicle_Type, dtype: category
Categories (16, object): ['Agricultural vehicle', 'Bus or coach (17 or more pass seats)', 'Car', 'Data missing or out of range', ..., 'Pedal cycle', 'Ridden horse', 'Taxi/Private hire car', 'Van / Goods 3.5 tonnes mgw or under']

Insight: According to the result, the most frequent/usual type of vehicle involve in accidents is car.

2. What is the usual light condition during the accidents?

In [22]:
ukroadaccident['Light_Conditions'].mode()
Out[22]:
0    Daylight
Name: Light_Conditions, dtype: category
Categories (5, object): ['Darkness - lighting unknown', 'Darkness - lights lit', 'Darkness - lights unlit', 'Darkness - no lighting', 'Daylight']

Insight: The result shows that accidents are usually happening during daylight.

3. In terms of road surface conditions, what is the the most frequent?

In [23]:
ukroadaccident['Road_Surface_Conditions'].mode()
Out[23]:
0    Dry
Name: Road_Surface_Conditions, dtype: category
Categories (6, object): ['Dry', 'Flood over 3cm. deep', 'Frost or ice', 'Snow', 'Wet or damp', 'unaccounted']

Insight: Based on the result, dry road surface condition are prone to accidents.

4. How many number of vehicles are mostly being involved during accidents?

In [24]:
ukroadaccident['Number_of_Vehicles'].mode()
Out[24]:
0    2
Name: Number_of_Vehicles, dtype: int64

Insight: The result shows that two vehicles are mostly involved during road accidents.

5. What type of road does the accidents usually happen?

In [25]:
ukroadaccident['Road_Type'].mode()
Out[25]:
0    Single carriageway
Name: Road_Type, dtype: category
Categories (6, object): ['Dual carriageway', 'One way street', 'Roundabout', 'Single carriageway', 'Slip road', 'unaccounted']

Insight: According to the result, accidents usually happen on a single carriageway road.

6. How many record of accidents per year?

In [26]:
ukroadaccident['Year'].value_counts()
Out[26]:
Year
2019    182115
2020    170591
2021    163554
2022    144419
Name: count, dtype: int64

Insight 1: The accidents record is gradually decreasing year by year from 27.57% in 2019 to 21.84% in 2022.

Insight 2: The highest number of accidents is on 2019, followed by 2020 at 25.83%, followed by 2021 with 24.76%, then 2022 is the lowest at 21.84%.

7. In terms of weather conditions, when does the least and most accidents record?

In [27]:
ukroadaccident['Weather_Conditions'].value_counts()
Out[27]:
Weather_Conditions
Fine no high winds       520885
Raining no high winds     79696
Other                     17150
unaccounted               14128
Raining + high winds       9615
Fine + high winds          8554
Snowing no high winds      6238
Fog or mist                3528
Snowing + high winds        885
Name: count, dtype: int64

Insight 1: Based on the result, the majority of incidents (79%) happen under fine weather with no high winds weather condition. So, even a fine weather is not safe from road accidents.

Insight 2: Raining without high winds condition accounts for 12%, which is the second most condition.

Insight 3: The other conditions are 2% to below 1% rate, and the lowest is snowy with high winds at 0.13%lowest is snowy with high winds at 0.13%

Insight 4: There is a small portion of incidents classified as “Other” or “Unaccounted” (4.7% combined) which indicates the records under unknown conditions.

~Aggregation~

8. What is the total number of accidents per road type?

In [28]:
ukroadaccident.groupby('Road_Type').size()
Out[28]:
Road_Type
Dual carriageway       99424
One way street         13559
Roundabout             43992
Single carriageway    492143
Slip road               7041
unaccounted             4520
dtype: int64

Insight 1: Single carriageways had the highest record of incidents with 74.55%, which means this type of road is accident-prone.

Insight 2: Dual carriageways is 15%, showing fewer incidents compared to single carriageways.

Insight 2: A very small portion of 0.7% is unaccounted, which is the lowest data but valuable still.

9. What is the average of number of casualties involved in accidents by accident severity?

In [29]:
ukroadaccident.groupby('Accident_Severity')['Number_of_Casualties'].mean()
Out[29]:
Accident_Severity
Fatal      1.903129
Serious    1.467280
Slight     1.331402
Name: Number_of_Casualties, dtype: float64

Insight: The average of accident severity in terms of fatal is 1.903129 (highest), serious is 1.467280, and slight is 1.331402 (lowest).

10. How do accident severities vary across different light conditions?

In [30]:
ukroadaccident.groupby(['Accident_Severity', 'Light_Conditions']).size().unstack()
Out[30]:
Light_Conditions Darkness - lighting unknown Darkness - lights lit Darkness - lights unlit Darkness - no lighting Daylight
Accident_Severity
Fatal 68 1860 45 1612 5076
Serious 794 19130 360 7174 60759
Slight 5622 108345 2138 28651 419045

Insight: We can conclude that all the three types of accident severity happens the most during the daylight, while the three of them also happens the least during darkness - lights unlit.

11. What is the distribution of accidents by road type and road surface condition?

In [31]:
ukroadaccident.groupby(['Road_Type', 'Road_Surface_Conditions']).size().unstack()
Out[31]:
Road_Surface_Conditions Dry Flood over 3cm. deep Frost or ice Snow Wet or damp unaccounted
Road_Type
Dual carriageway 66205 302 2408 909 29533 67
One way street 10068 5 183 76 3195 32
Roundabout 30698 22 745 232 12209 86
Single carriageway 332698 672 14918 4585 138743 527
Slip road 4714 9 178 60 2074 6
unaccounted 3438 7 85 28 954 8

Insight: The result shows that all of the road types had the same highest number of accidents in terms of road surface conditions which is dry, they also had the same least number of accidents which is flood over 3cm. deep (road surface condition).

12. What type of area does accidents happen the most?

In [32]:
ukroadaccident.groupby('Urban_or_Rural_Area').size()
Out[32]:
Urban_or_Rural_Area
Rural          238990
Unallocated        11
Urban          421678
dtype: int64

Insight 1: According to the result, accidents usually happen at Urban areas with 63.82%. On the other hand, Rural areas account for 36%, still a significant portion.

Insight 2: The Rural accidents record somehow half of the records in Urban areas, both are significant because if we only look at one a time, both are large numbers.

13. What month does accidents usually happens?

In [33]:
ukroadaccident.groupby('Month').size()
Out[33]:
Month
April        51744
August       53913
December     51836
February     49491
January      52872
July         57445
June         56481
March        54086
May          56352
November     60424
October      59580
September    56455
dtype: int64

Insight 1: The result shows that accidents usually happen during the month of November with 9.15% and February (7.49%) has the lowest.

Insight 2: Most months are fairly close, around 7.5–8.7%, showing fairly consistent incident distribution across the year.

Insight 3: Summer months (June–August) account for roughly ~25% combined, indicating higher traffic or activity during that period.

Insight 4: The slight peaks in October(9.02%)–November(9.15%) could indicate weather changes, holidays, or increased traffic contributing to incidents.

14. What is the distribution of accidents by accident severity and vehicle type?

In [34]:
ukroadaccident.groupby(['Accident_Severity','Vehicle_Type']).size().unstack()
Out[34]:
Vehicle_Type Agricultural vehicle Bus or coach (17 or more pass seats) Car Data missing or out of range Goods 7.5 tonnes mgw and over Goods over 3.5t. and under 7.5t Minibus (8 - 16 passenger seats) Motorcycle 125cc and under Motorcycle 50cc and under Motorcycle over 125cc and up to 500cc Motorcycle over 500cc Other vehicle Pedal cycle Ridden horse Taxi/Private hire car Van / Goods 3.5 tonnes mgw or under
Accident_Severity
Fatal 21 325 6577 0 216 67 29 189 95 105 339 70 6 0 155 467
Serious 282 3373 66461 0 2321 857 276 2031 1014 1014 3457 767 39 0 1771 4554
Slight 1644 22180 424954 6 14770 5172 1671 13049 6494 6537 21861 4800 152 4 11368 29139

Insight: According to the result, car is the vehicle type that got the highest accident of all the three types of accident severity, while ridden horse has the least.

~Correlation~

15. Is there a correlation between the number of casualties and number of vehicles?

In [35]:
ukroadaccident['Number_of_Casualties'].corr(ukroadaccident['Number_of_Vehicles'])
Out[35]:
np.float64(0.2288888612692756)

Insight: The result shows that number of casualties and number of vehicles has no correlation.

16. Is there a correlation betweem latitude number of casualties?

In [36]:
ukroadaccident['Latitude'].corr(ukroadaccident['Number_of_Casualties'])
Out[36]:
np.float64(0.032200686625906395)

Insight: Based on the result, there is no correlation between laltitude and number of casualties.

17. Is there a correlation between longitude and number of casualties?

In [37]:
ukroadaccident['Longitude'].corr(ukroadaccident['Number_of_Casualties'])
Out[37]:
np.float64(-0.0404056457884545)

Insight: The result shows that there is no correlation between longitude and number of casualties.

18. Is there a significant difference between accident severity and number of vehicles?

In [38]:
ukroadaccident['Accident_Severity'].unique()
Out[38]:
['Serious', 'Slight', 'Fatal']
Categories (3, object): ['Fatal', 'Serious', 'Slight']
In [39]:
vehiserious = ukroadaccident[ukroadaccident['Accident_Severity'] == 'Serious']['Number_of_Vehicles']
vehislight = ukroadaccident[ukroadaccident['Accident_Severity'] == 'Slight']['Number_of_Vehicles']
vehifatal = ukroadaccident[ukroadaccident['Accident_Severity'] == 'Fatal']['Number_of_Vehicles']
In [40]:
result, pvalue = f_oneway(vehiserious, vehislight, vehifatal)
pvalue
Out[40]:
np.float64(0.0)

Insight: The result shows that there is extremely significance between the accident severity and number of vehicles, it shows that they are very related to each other.

19. Is there a significant difference between the area and number of casualties?

In [41]:
ukroadaccident['Urban_or_Rural_Area'].unique()
Out[41]:
['Urban', 'Rural', 'Unallocated']
Categories (3, object): ['Rural', 'Unallocated', 'Urban']
In [42]:
urbancasualty = ukroadaccident[ukroadaccident['Urban_or_Rural_Area'] == 'Urban']['Number_of_Casualties']
ruralcasualty = ukroadaccident[ukroadaccident['Urban_or_Rural_Area'] == 'Rural']['Number_of_Casualties']
unallocatedcasualty = ukroadaccident[ukroadaccident['Urban_or_Rural_Area'] == 'Unallocated']['Number_of_Casualties']
In [43]:
result, pvalue = f_oneway(urbancasualty, ruralcasualty, unallocatedcasualty)
pvalue
Out[43]:
np.float64(0.0)

Insight: According to the result, there is a significant difference between the areas and the number of casualties.

20. Is there a significant difference between light conditions and latitude?

In [44]:
ukroadaccident['Light_Conditions'].unique()
Out[44]:
['Darkness - lights lit', 'Daylight', 'Darkness - lighting unknown', 'Darkness - lights unlit', 'Darkness - no lighting']
Categories (5, object): ['Darkness - lighting unknown', 'Darkness - lights lit', 'Darkness - lights unlit', 'Darkness - no lighting', 'Daylight']
In [45]:
lighta = ukroadaccident[ukroadaccident['Light_Conditions'] == 'Darkness - lights lit']['Latitude']
lightb = ukroadaccident[ukroadaccident['Light_Conditions'] == 'Daylight']['Latitude']
lightc = ukroadaccident[ukroadaccident['Light_Conditions'] == 'Darkness - lighting unknown']['Latitude']
lightd = ukroadaccident[ukroadaccident['Light_Conditions'] == 'Darkness - lights unlit']['Latitude']
lighte = ukroadaccident[ukroadaccident['Light_Conditions'] == 'Darkness - no lighting']['Latitude']
In [46]:
result, pvalue = f_oneway(lighta, lightb, lightc, lightd, lighte)
pvalue
Out[46]:
np.float64(1.224998791423201e-27)

Insight: The result shows that the difference between light conditions and latitude is very significant, mean the light conditions has an extreme effect on the latitude.

---Additional Five Questions about Accident Date---

-UniVariate-

21. Which day of the week had the most accidents record?

In [47]:
ukroadaccident['DayofWeek'].value_counts()
Out[47]:
DayofWeek
5    107178
2     99558
3     99511
4     97900
1     94550
6     89302
0     72680
Name: count, dtype: int64

Insight: Based on the result, accidents mostly occurs on Saturday, and accidents happen the least during Mondays.

-BiVariate-

22. What is the number of accidents per month in each year?

In [48]:
accidentpermonth = ukroadaccident.groupby(['Month', 'Year']).size()
accidentpermonth.unstack()
Out[48]:
Year 2019 2020 2021 2022
Month
April 14125 13394 12715 11510
August 15044 13366 13415 12088
December 14708 13794 13709 9625
February 13253 14353 10950 10935
January 15355 14133 13417 9967
July 15862 14630 14300 12653
June 15528 14205 13936 12812
March 15049 13494 13202 12341
May 15833 14336 13811 12372
November 16559 14770 15473 13622
October 15528 15684 14834 13534
September 15271 14432 13792 12960

Insight 1: Based on their percentage equivalence, November is consistenly the peak month of accidents across all four years with two consequtive above 9% rate (2021, and 2022). Hoevwer, October also has high percentage with three ratings of 9% and above but November slightly beats it in most years.

Insight 2: February has the lowest percentage of accidents especially 2021 at just 6.7%. It suggests that road accidents tend to be more common in October, while February is generally a safer month.

Insight 3: Most of the months had high road accident rate on the year 2019, and we could see from the numbers that on the years 2020-2022, the number had decreased and we can conclude that this is perhaps because of the pandemic.

23. What is the state of accident severity across years?

In [49]:
ukroadaccident.groupby(['Accident_Severity', 'Year']).size().unstack()
Out[49]:
Year 2019 2020 2021 2022
Accident_Severity
Fatal 2714 2341 2057 1549
Serious 24322 23121 21997 18777
Slight 155079 145129 139500 124093

Insight 1: From the number itself, it shows that deadly crashes are rare compared to other types of accident severity. Fatal accidents are only around 1% of all cases. They even decreased slightly over the years, from 1.49% in 2019 down to 1.07% in 2022.

Insight 2: On the other hand, slight accidents had the highest consistent record at around 85% every year.

Insight 3: Seious accidents are in the middle of the percentage, staying between 13–14%.

Insight 4: Accidents across the years were slight and less harmful, a small portion are serious, and very few cases are fatal.

-MultiVariate-

24. What is the average casualties per month, year, and road surface conditions?

In [50]:
ukroadaccident.groupby(['Month', 'Year', 'Road_Surface_Conditions'])['Number_of_Casualties'].mean().unstack()
Out[50]:
Road_Surface_Conditions Dry Flood over 3cm. deep Frost or ice Snow Wet or damp unaccounted
Month Year
April 2019 1.360358 1.000000 1.200000 NaN 1.385888 1.000000
2020 1.339197 1.538462 1.560606 1.442308 1.442780 1.000000
2021 1.353704 1.000000 1.000000 2.000000 1.420137 1.181818
2022 1.357578 1.000000 1.500000 1.250000 1.485714 1.181818
August 2019 1.383017 1.636364 1.000000 1.000000 1.458948 1.304348
2020 1.375027 1.578947 1.000000 2.000000 1.450013 1.304348
2021 1.385929 2.090909 NaN 1.000000 1.463757 1.000000
2022 1.379569 1.750000 1.500000 2.000000 1.415216 1.400000
December 2019 1.286649 1.531915 1.350850 1.181818 1.422541 1.000000
2020 1.335708 1.708333 1.345857 1.252427 1.401117 1.142857
2021 1.274566 1.500000 1.380342 1.374680 1.388427 1.461538
2022 1.297656 1.600000 1.385561 1.376316 1.415217 1.125000
February 2019 1.315298 1.083333 1.287305 1.312000 1.401046 1.133333
2020 1.318313 1.600000 1.346648 1.661017 1.383009 1.071429
2021 1.326901 1.176471 1.364519 1.303318 1.362491 1.272727
2022 1.325941 1.714286 1.317593 1.402632 1.388079 1.187500
January 2019 1.301965 1.358974 1.331887 1.278571 1.403802 1.150000
2020 1.305425 1.515625 1.308008 1.301887 1.369336 1.217391
2021 1.331680 1.333333 1.315061 1.346535 1.387118 1.357143
2022 1.285207 1.111111 1.336806 1.334642 1.347077 1.222222
July 2019 1.342417 1.538462 1.000000 1.400000 1.459278 1.444444
2020 1.324689 1.583333 1.000000 1.000000 1.445199 1.500000
2021 1.340411 1.444444 1.000000 NaN 1.474479 1.066667
2022 1.339222 1.750000 1.000000 1.000000 1.470771 1.000000
June 2019 1.336180 1.506667 1.000000 1.000000 1.447362 1.117647
2020 1.333754 1.166667 1.000000 NaN 1.416375 1.076923
2021 1.329111 1.555556 NaN 1.333333 1.442604 1.714286
2022 1.339122 2.000000 1.000000 NaN 1.399510 1.000000
March 2019 1.330359 1.294118 1.231579 1.411111 1.435946 1.062500
2020 1.333562 1.352941 1.452555 1.352381 1.405273 1.450000
2021 1.334661 1.200000 1.346939 1.678571 1.413530 1.250000
2022 1.325219 1.333333 1.246914 1.500000 1.411477 1.227273
May 2019 1.346532 1.684211 1.000000 1.000000 1.442656 1.100000
2020 1.337205 1.400000 1.500000 1.333333 1.457971 1.352941
2021 1.354493 1.666667 1.500000 1.333333 1.417683 1.071429
2022 1.338688 1.250000 1.333333 1.000000 1.500753 1.250000
November 2019 1.290865 1.260870 1.369295 1.428571 1.400236 1.043478
2020 1.300952 1.583333 1.434132 1.384615 1.387696 1.200000
2021 1.294445 1.452381 1.243243 1.923077 1.393972 1.421053
2022 1.318023 1.444444 1.356423 1.371429 1.389572 1.142857
October 2019 1.330498 1.272727 1.400000 1.000000 1.393419 1.153846
2020 1.314318 1.518519 1.347150 1.485714 1.397198 1.090909
2021 1.324756 1.533333 2.200000 1.000000 1.426586 1.076923
2022 1.336587 1.705882 1.188679 1.714286 1.388816 1.210526
September 2019 1.339429 1.250000 1.000000 1.500000 1.435417 1.307692
2020 1.323790 1.509804 1.000000 1.000000 1.395795 1.083333
2021 1.328964 2.111111 NaN 1.500000 1.401223 1.411765
2022 1.332019 1.400000 1.000000 2.000000 1.408525 1.636364

Insight 1: The average casualties are slightly higher on wet/damp roads (around 1.40–1.47) compared to dry roads (around 1.30–1.36) across almost months and years.

Insight 2: The "Flood over 3cm deep" condition often shows averages above 1.5 and sometimes above 2.0 casualties per accident.

Averages of frost/ice and snow condition variesfrost/ice and snow condition varies a lot, both of them has usually around 1.0 average but there are times that it spikes to above 2.0 (October 2021 has the highest value with 2.2).

Insight 4: Most accidents happen on dry roads but accidents on wet and flooded roads are a bit more dangerous. Snow and ice road surface condition tend to happen less frequent but it is the most dangerous and causes more casualties when they do happen.

25. What is the distribution of accidents by year, month, and rural/urban area?

In [51]:
ukroadaccident.groupby(['Urban_or_Rural_Area','Year', 'Month']).size().unstack()
Out[51]:
Month April August December February January July June March May November October September
Urban_or_Rural_Area Year
Rural 2019 5169 6001 5499 4874 5807 6047 5835 5350 5802 5974 5518 5682
2020 4752 5204 5167 5226 5182 5415 5105 4947 5184 5251 5415 5143
2021 4555 5256 5106 4144 4925 5277 4943 4391 4873 5456 5260 4830
2022 3956 4537 3589 3896 3626 4476 4317 4118 4266 4671 4593 4380
Unallocated 2019 0 0 2 0 1 1 1 0 0 0 0 0
2020 1 0 1 0 0 0 1 1 2 0 0 0
2021 0 0 0 0 0 0 0 0 0 0 0 0
2022 0 0 0 0 0 0 0 0 0 0 0 0
Urban 2019 8956 9043 9207 8379 9547 9814 9692 9699 10031 10585 10010 9589
2020 8641 8162 8626 9127 8951 9215 9099 8546 9150 9519 10269 9289
2021 8160 8159 8603 6806 8492 9023 8993 8811 8938 10017 9574 8962
2022 7554 7551 6036 7039 6341 8177 8495 8223 8106 8951 8941 8580

Insight 1: In all months and years, accidents mostly happens in urban areas at 76%-80%. While accidents in rural areas are lower with 19%-23%.

Insight 2: The highest percentage in urban areas was in November 2021 at 80.16%, followed by both September 2021 and October 2020 at 80.15%. While the lowest was April 2019 (76.56%).

Insight 3: In rural areas, the lowest percentage was 19.84% on November 2021, while the highest was on April 2019 at 23.44%.

Insight 4: Most accidents happen in urban areas (cities) regardless of the month and year. Rural accidents are fewer but the still have a peak month (April).

---Additional questions with matplotlib---

In [52]:
import matplotlib.pyplot as plt
In [53]:
ukroadaccident.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 660679 entries, 0 to 660678
Data columns (total 19 columns):
 #   Column                   Non-Null Count   Dtype         
---  ------                   --------------   -----         
 0   Index                    660679 non-null  category      
 1   Accident_Severity        660679 non-null  category      
 2   Accident Date            660679 non-null  datetime64[ns]
 3   Latitude                 660679 non-null  float64       
 4   Light_Conditions         660679 non-null  category      
 5   District Area            660679 non-null  category      
 6   Longitude                660679 non-null  float64       
 7   Number_of_Casualties     660679 non-null  int64         
 8   Number_of_Vehicles       660679 non-null  int64         
 9   Road_Surface_Conditions  660679 non-null  category      
 10  Road_Type                660679 non-null  category      
 11  Urban_or_Rural_Area      660679 non-null  category      
 12  Weather_Conditions       660679 non-null  category      
 13  Vehicle_Type             660679 non-null  category      
 14  Year                     660679 non-null  category      
 15  Month_Number             660679 non-null  int32         
 16  Month                    660679 non-null  category      
 17  Day                      660679 non-null  int32         
 18  DayofWeek                660679 non-null  int32         
dtypes: category(11), datetime64[ns](1), float64(2), int32(3), int64(2)
memory usage: 61.6 MB

26. What is the accident distribution by severity at Birmingham (highest accident record district area)?

In [54]:
birmingham = ukroadaccident[ukroadaccident['District Area'] == 'Birmingham']
accidentbirmingham = birmingham.groupby(['Accident_Severity', 'Light_Conditions']).size()
accidentbirmingham
Out[54]:
Accident_Severity  Light_Conditions           
Fatal              Darkness - lighting unknown       2
                   Darkness - lights lit            47
                   Darkness - lights unlit           3
                   Darkness - no lighting            0
                   Daylight                         53
Serious            Darkness - lighting unknown       2
                   Darkness - lights lit           483
                   Darkness - lights unlit           3
                   Darkness - no lighting            6
                   Daylight                        980
Slight             Darkness - lighting unknown      78
                   Darkness - lights lit          3142
                   Darkness - lights unlit          45
                   Darkness - no lighting           13
                   Daylight                       8634
dtype: int64
In [55]:
accidentbirmingham.plot(kind='bar', stacked=True, figsize=(10,6))
plt.title("Accidents in Birmingham by Severity & Light Conditions")
plt.xlabel("Light Conditions")
plt.ylabel("Number of Accidents")
plt.legend(title="Accident Severity")
plt.show()
No description has been provided for this image

Insight: The illustration shows that most slight severity accidents happen during daylight in Birmingham. Followed by slight accidents in darkness- light lit. Lastly, there's a small portion of serious accidents during daylight.

27. What is the car accident distribution by severity per year and month?

In [56]:
car = ukroadaccident[ukroadaccident['Vehicle_Type'] == 'Car']
grouped = car.groupby(['Year', 'Accident_Severity']).size().reset_index(name='Count')
grouped.columns = ['Year', 'Accident_Severity', 'Count']
grouped
Out[56]:
Year Accident_Severity Count
0 2019 Fatal 1948
1 2019 Serious 17469
2 2019 Slight 110887
3 2020 Fatal 1772
4 2020 Serious 17384
5 2020 Slight 108738
6 2021 Fatal 1616
7 2021 Serious 17143
8 2021 Slight 109267
9 2022 Fatal 1241
10 2022 Serious 14465
11 2022 Slight 96062
In [57]:
pivoted = grouped.pivot(index='Year', columns='Accident_Severity', values='Count')
In [58]:
pivoted.plot(kind='line', marker='o', figsize=(10,6))
plt.title("Car Accident Distribution by Severity per Year")
plt.xlabel("Year")
plt.ylabel("Number of Car Accidents")
plt.legend(title="Severity")
plt.show()
No description has been provided for this image

Insight: As we can see from the line graph, there is a huge gap between slight severity and the other two severities. Fatality rate is the lowest of them all. Serious car accidents are around 15,000 number of accidents. While slight severity car accidents are on the top with above 100k records across 2019-2021, and it dropped a bit in 2022.

28. What is the accident distribution by road surface conditions per areas (rural, urban)?

In [59]:
surfacearea = ukroadaccident.groupby(['Urban_or_Rural_Area', 'Road_Surface_Conditions']).size().unstack()
surfacearea.plot(kind='bar', figsize=(10,6))
plt.title("Accidents by Road Surface Conditions and Area Type")
plt.ylabel("Number of Accidents")
plt.xlabel("Area (Urban / Rural)")
plt.legend(title="Road Surface Conditions")
plt.show()
No description has been provided for this image

Insight: Both rural and urban have both types of condition rate in terms of road surface. However, Urban got the highest number of accidents on dry or wet/damp surface conditions and on the other hand, Rural is much higher in terms of number of accidents on frost/ice and snow condition.

29. Comparing the weather condtions among the three highest disctrict areas and lowest area.

In [60]:
ukroadaccident['District Area'].value_counts().head(3)
Out[60]:
District Area
Birmingham    13491
Leeds          8898
Manchester     6720
Name: count, dtype: int64
In [61]:
ukroadaccident['District Area'].value_counts().tail(1)
Out[61]:
District Area
Clackmannanshire    91
Name: count, dtype: int64
In [62]:
weatherbirmingham = ukroadaccident[ukroadaccident['District Area'] == 'Birmingham']
birmiweather = weatherbirmingham['Weather_Conditions'].value_counts()
birmiweather.plot(kind='line', marker='o', figsize=(10,6))
plt.title("Weather condition in Birmingham")
plt.xlabel("Weather Conditions")
plt.ylabel("Number of Accidents")
plt.xticks(rotation=45)
plt.show()
No description has been provided for this image
In [63]:
weatherleeds = ukroadaccident[ukroadaccident['District Area'] == 'Leeds']
leedsweather = weatherleeds['Weather_Conditions'].value_counts()
leedsweather.plot(kind='line', marker='o', figsize=(10,6))
plt.title("Weather condition in Leeds")
plt.xlabel("Weather Conditions")
plt.ylabel("Number of Accidents")
plt.xticks(rotation=45)
plt.show()
No description has been provided for this image
In [64]:
weathermanchester = ukroadaccident[ukroadaccident['District Area'] == 'Manchester']
manchesterweather = weathermanchester['Weather_Conditions'].value_counts()
manchesterweather.plot(kind='line', marker='o', figsize=(10,6))
plt.title("Weather condition in Manchester")
plt.xlabel("Weather Conditions")
plt.ylabel("Number of Accidents")
plt.xticks(rotation=45)
plt.show()
No description has been provided for this image
In [65]:
weatherclack = ukroadaccident[ukroadaccident['District Area'] == 'Clackmannanshire']
clackweather = weatherclack['Weather_Conditions'].value_counts()
clackweather.plot(kind='line', marker='o', figsize=(10,6))
plt.title("Weather condition in Clackmannanshire")
plt.xlabel("Weather Conditions")
plt.ylabel("Number of Accidents")
plt.xticks(rotation=45)
plt.show()
No description has been provided for this image

Insight 1: It looks like there isnt much difference but if we really analyze it we can see their difference.

Insight 2: First, all four districts has the same type of weather condtion (free no high winds) as the highest number of incidents(regardless of the total number of incidents per district) , and "snowy without winds" is the lowest.

Insight 3: Raining without high winds is the second most condition among the four districts too.

Insight 4: From the third to eighth conditions across all charts now varies.

In [66]:
#look for the data without the the outlier(fine no high winds) in all four district above.

---SIX DISTRICT AREA ANALYSIS COMPARED TO NATIONAL TRENDS---

In [67]:
import folium
from folium.plugins import HeatMap

-National Analysis-

In [68]:
import matplotlib.pyplot as plt

# 1. Group by Year FIRST, then Month (critical order!)
monthly_data = ukroadaccident.groupby(['Year', 'Month']).size()

# 2. Unstack Year to create separate lines for each year
monthly_data = monthly_data.unstack(level=0)

# 3. Plot with clean formatting (only 4 lines!)
monthly_data.plot(
    marker='o', 
    figsize=(10, 5),
    title='Monthly Road Accidents by Year (2019-2022)'
)
plt.xticks(range(0, 12), ['Jan','Feb','Mar','Apr','May','Jun','Jul','Aug','Sep','Oct','Nov','Dec'])
plt.legend(title='Year and Month Severity Comparison', bbox_to_anchor = (1.05,1), loc='upper left')
plt.grid(alpha=0.3)
plt.show()
No description has been provided for this image

-National Seasonal Analysis-

In [69]:
# Accidents per year
accidents_per_year = ukroadaccident['Year'].value_counts().sort_index()
accidents_per_month = ukroadaccident['Month_Number'].value_counts().sort_index()
accidents_per_day = ukroadaccident['DayofWeek'].value_counts().sort_index()


#PER YEAR 
accidents_per_year.plot(kind='bar')
plt.title('Accidents per Year')
plt.xlabel('Year')
plt.ylabel('Number of Accidents')
plt.show()

month_names = {
    1: 'January', 2: 'February', 3: 'March', 4: 'April',
    5: 'May', 6: 'June', 7: 'July', 8: 'August',
    9: 'September', 10: 'October', 11: 'November', 12: 'December'
}

# Replace numerical months with their names
accidents_per_month.index = accidents_per_month.index.astype(int).map(month_names)


accidents_per_month.plot(kind='bar')
plt.title('Accidents per Month')
plt.xlabel('Month')
plt.ylabel('Number of Accidents')
plt.show()


#accident per day
day_names = {
    0: 'Monday',
    1: 'Tuesday',
    2: 'Wednesday',
    3: 'Thursday',
    4: 'Friday',
    5: 'Saturday',
    6: 'Sunday'
}

accidents_per_day.index = accidents_per_day.index.map(day_names)
accidents_per_day.plot(kind='bar')
print(accidents_per_day)
No description has been provided for this image
No description has been provided for this image 
DayofWeek
Monday        72680
Tuesday       94550
Wednesday     99558
Thursday      99511
Friday        97900
Saturday     107178
Sunday        89302
Name: count, dtype: int64
No description has been provided for this image

Accident Severity

In [70]:
severity_counts = ukroadaccident['Accident_Severity'].value_counts()

plt.pie(severity_counts,
        labels = severity_counts.index,
        autopct = '%1.1f%%',
        startangle = 90)
plt.gcf().set_size_inches(10,7)
plt.title('Accident Severity Distribution')
plt.show()

severity_2019 = ukroadaccident[ukroadaccident['Year'] == 2019]
# severity_2019
sc_2019 = severity_2019['Accident_Severity'].value_counts()

plt.pie(sc_2019,
        labels = sc_2019.index,
        autopct = '%1.1f%%',
        startangle = 90)
plt.gcf().set_size_inches(10,7)
plt.title('Accident Severity Distribution')
plt.show()
No description has been provided for this image
No description has been provided for this image
In [140]:
import folium
from folium.plugins import HeatMap

# ukmap = folium.Map(location=[54, -2], zoom_start=6)

# for idx, row in ukroadaccident.iterrows():
#     folium.Marker([row['Latitude'], row['Longitude']]).add_to(ukmap)

# ukmap.save('Accident_map.html')

uk = list(zip(ukroadaccident['Latitude'], ukroadaccident['Longitude']))
m = folium.Map(location=[ukroadaccident['Latitude'].mean(), ukroadaccident['Longitude'].mean()], zoom_start=10)

HeatMap(uk).add_to(m)
m
# m.save('Aberdeen_heatmap.html')
Out[140]:
Make this Notebook Trusted to load map: File -> Trust Notebook
In [72]:
ukroadaccident['District Area'].unique()
Out[72]:
['Kensington and Chelsea', 'Westminster', 'Richmond upon Thames', 'Hammersmith and Fulham', 'Hounslow', ..., 'County Durham', 'Shropshire', 'Central Bedfordshire', 'Cornwall', 'Wiltshire']
Length: 422
Categories (422, object): ['Aberdeen City', 'Aberdeenshire', 'Adur', 'Allerdale', ..., 'Wycombe', 'Wyre', 'Wyre Forest', 'York']

Using District Area to make a comparison to a general result

1. Aberdeen City Analysis

In [73]:
aberdeen_df = ukroadaccident[ukroadaccident['District Area'] == 'Aberdeen City']
In [74]:
monthly_aberdeendata = aberdeen_df.groupby(['Year', 'Month']).size()

monthly_aberdeendata = monthly_aberdeendata.unstack(level=0)

monthly_aberdeendata.plot(
    marker='o', 
    figsize=(10, 5),
    title='Monthly Road Accidents in Aberdeen City by Year (2019-2022)'
)
plt.xticks(range(0, 12), ['Jan','Feb','Mar','Apr','May','Jun','Jul','Aug','Sep','Oct','Nov','Dec'])
plt.legend(title='Yearly Accident Comparison', bbox_to_anchor = (1.05,1), loc='upper left')
plt.grid(alpha=0.3)
plt.show()
No description has been provided for this image

Insight 1: On the national analytics, there is a noticeable spike in November 2019. However in Aberdeen City, there was a peak in February 2020. The lowest point on the national record was in December 2022, while in Aberdeen City there was a consistent zero record of all months on 2022. Overall, the national record is generally decreasing by year, but in the Aberdeen City it is quite unstable by year.

In [75]:
# Accidents per year
accidents_per_year = aberdeen_df['Year'].value_counts().sort_index()
accidents_per_month = aberdeen_df['Month_Number'].value_counts().sort_index()
accidents_per_day = aberdeen_df['DayofWeek'].value_counts().sort_index()


#PER YEAR 
accidents_per_year.plot(kind='bar')
plt.title('Accidents per Year')
plt.xlabel('Year')
plt.ylabel('Number of Accidents')
plt.show()

month_names = {
    1: 'January', 2: 'February', 3: 'March', 4: 'April',
    5: 'May', 6: 'June', 7: 'July', 8: 'August',
    9: 'September', 10: 'October', 11: 'November', 12: 'December'
}

# Replace numerical months with their names
accidents_per_month.index = accidents_per_month.index.map(month_names)


accidents_per_month.plot(kind='bar')
plt.title('Accidents per Month')
plt.xlabel('Month')
plt.ylabel('Number of Accidents')
plt.show()

#accident per day
day_names = {
    0: 'Monday',
    1: 'Tuesday',
    2: 'Wednesday',
    3: 'Thursday',
    4: 'Friday',
    5: 'Saturday',
    6: 'Sunday'
}

accidents_per_day.index = accidents_per_day.index.map(day_names)
accidents_per_day.plot(kind='bar')
print(accidents_per_day)
No description has been provided for this image
No description has been provided for this image 
DayofWeek
Monday       125
Tuesday      189
Wednesday    212
Thursday     205
Friday       204
Saturday     210
Sunday       178
Name: count, dtype: int64
No description has been provided for this image

Insight 2: In terms of Year, the sequence of the national records is 2019 -being the highest- to 2020 to 2021 and to 2022 (lowest). While, in Aberdeen City, they got 2020 with the higest record to 2021 to 2019 and to 2020 with zero record.

Insight 3: In terms of Month, accidents usually happen on November with almost 60,000 records, while February is the least month with almost 50,000 records (national record). On the other hand, road accidents in the Aberdeen City usually happen on the month of May with above 120 records, while the least accidents is on the month of April with 90 records.

Insight 4: In terms of Day of the Week, Saturday is the day with highest records (above 100,000) while Monday is the lowest (around 70,000). While in Aberdeen City, accidents usually happen during Wednesday (around 200), and the least is on Monday (125 records). Though most of the weekdays have the most records.

Accident Severity in Aberdeen City

In [76]:
severity_aberdeen  = aberdeen_df['Accident_Severity'].value_counts()

plt.pie(severity_aberdeen,
        labels = severity_aberdeen.index,
        autopct = '%1.1f%%',
        startangle = 90)
plt.gcf().set_size_inches(10,7)
plt.title('Accident Severity Distribution')
plt.show()

severity_2019_aberdeen = aberdeen_df[aberdeen_df['Year'] == 2019]
# severity_2019
lsc_2019 = severity_2019_aberdeen['Accident_Severity'].value_counts()

plt.pie(lsc_2019,
        labels = lsc_2019.index,
        autopct = '%1.1f%%',
        startangle = 90)
plt.gcf().set_size_inches(10,7)
plt.title('Accident Severity Distribution')
plt.show()
No description has been provided for this image
No description has been provided for this image

Insight 5: Both records of the National and the Aberdeen City is almost the same with slight severity being the highest and serious severity being the second, and the lowest is the 'fatal'.

In [136]:
aberdeen_df = ukroadaccident[ukroadaccident['District Area'] == 'Aberdeen City']
aber = list(zip(aberdeen_df['Latitude'], aberdeen_df['Longitude']))
m = folium.Map(location=[aberdeen_df['Latitude'].mean(), aberdeen_df['Longitude'].mean()], zoom_start=10)

HeatMap(aber).add_to(m)
# m.save('Aberdeen_heatmap.html')
m
Out[136]:
Make this Notebook Trusted to load map: File -> Trust Notebook

Wyre Analysis

In [78]:
ukroadaccident['District Area'].unique()
Out[78]:
['Kensington and Chelsea', 'Westminster', 'Richmond upon Thames', 'Hammersmith and Fulham', 'Hounslow', ..., 'County Durham', 'Shropshire', 'Central Bedfordshire', 'Cornwall', 'Wiltshire']
Length: 422
Categories (422, object): ['Aberdeen City', 'Aberdeenshire', 'Adur', 'Allerdale', ..., 'Wycombe', 'Wyre', 'Wyre Forest', 'York']
In [79]:
wyre_df = ukroadaccident[ukroadaccident['District Area'] == 'Wyre']
In [80]:
monthly_wyredata = wyre_df.groupby(['Year', 'Month']).size()

monthly_wyredata = monthly_wyredata.unstack(level=0)

monthly_wyredata.plot(
    marker='o', 
    figsize=(10, 5),
    title='Monthly Road Accidents in Wyre by Year (2019-2022)'
)
plt.xticks(range(0, 12), ['Jan','Feb','Mar','Apr','May','Jun','Jul','Aug','Sep','Oct','Nov','Dec'])
plt.legend(title='Yearly Accident Comparison', bbox_to_anchor = (1.05,1), loc='upper left')
plt.grid(alpha=0.3)
plt.show()
No description has been provided for this image

Insight 1: Compared to the national analytics, Wyre had its peak in August 2021, and the lowest point in December 2022. The distribution of accidents by year and month is quite unstable in Wyre.

In [81]:
# Accidents per year
accidents_per_year = wyre_df['Year'].value_counts().sort_index()
accidents_per_month = wyre_df['Month_Number'].value_counts().sort_index()
accidents_per_day = wyre_df['DayofWeek'].value_counts().sort_index()


#PER YEAR 
accidents_per_year.plot(kind='bar')
plt.title('Accidents per Year')
plt.xlabel('Year')
plt.ylabel('Number of Accidents')
plt.show()

month_names = {
    1: 'January', 2: 'February', 3: 'March', 4: 'April',
    5: 'May', 6: 'June', 7: 'July', 8: 'August',
    9: 'September', 10: 'October', 11: 'November', 12: 'December'
}

# Replace numerical months with their names
accidents_per_month.index = accidents_per_month.index.map(month_names)


accidents_per_month.plot(kind='bar')
plt.title('Accidents per Month')
plt.xlabel('Month')
plt.ylabel('Number of Accidents')
plt.show()

#accident per day
day_names = {
    0: 'Monday',
    1: 'Tuesday',
    2: 'Wednesday',
    3: 'Thursday',
    4: 'Friday',
    5: 'Saturday',
    6: 'Sunday'
}

accidents_per_day.index = accidents_per_day.index.map(day_names)
accidents_per_day.plot(kind='bar')
print(accidents_per_day)
No description has been provided for this image
No description has been provided for this image 
DayofWeek
Monday       133
Tuesday      163
Wednesday    193
Thursday     182
Friday       188
Saturday     197
Sunday       182
Name: count, dtype: int64
No description has been provided for this image

Insight 2: In terms of Year, the national record is decreasing by year. While in Wyre, year 2019 and 2021, and year 2020 and 2022, almost have the same record, it looks like it has a pattern.

Insight 3: In terms of Month, both has the month of November with being the highest. February is the lowest month in the national analytics while in Wyre, the lowest is the month of September.

Insight 4: In terms of Day of the Week, Saturday is the day with highest records in both of them (national record and Wyre), though Wednesday is almost the same as Saturday, then same goes with the lowest which is also both Monday.

Wyre Accident Severity

In [82]:
severity_wyre  = wyre_df['Accident_Severity'].value_counts()

plt.pie(severity_wyre,
        labels = severity_wyre.index,
        autopct = '%1.1f%%',
        startangle = 90)
plt.gcf().set_size_inches(10,7)
plt.title('Accident Severity Distribution')
plt.show()

severity_2019_wyre = wyre_df[wyre_df['Year'] == 2019]
# severity_2019
lsc_2019 = severity_2019_wyre['Accident_Severity'].value_counts()

plt.pie(lsc_2019,
        labels = lsc_2019.index,
        autopct = '%1.1f%%',
        startangle = 90)
plt.gcf().set_size_inches(10,7)
plt.title('Accident Severity Distribution')
plt.show()
No description has been provided for this image
No description has been provided for this image

Insight 5: Both also has the same distribution of accident severity in the graph, but the record of fatality in Wyre is slightly higher than the usual.

In [137]:
wyre_df = ukroadaccident[ukroadaccident['District Area'] == 'Wyre']
wyre = list(zip(wyre_df['Latitude'], wyre_df['Longitude']))
m = folium.Map(location=[wyre_df['Latitude'].mean(), wyre_df['Longitude'].mean()], zoom_start=10)

HeatMap(wyre).add_to(m)
# m.save('Wyre_heatmap.html')
m
Out[137]:
Make this Notebook Trusted to load map: File -> Trust Notebook
In [84]:
ukroadaccident['District Area'].unique()
Out[84]:
['Kensington and Chelsea', 'Westminster', 'Richmond upon Thames', 'Hammersmith and Fulham', 'Hounslow', ..., 'County Durham', 'Shropshire', 'Central Bedfordshire', 'Cornwall', 'Wiltshire']
Length: 422
Categories (422, object): ['Aberdeen City', 'Aberdeenshire', 'Adur', 'Allerdale', ..., 'Wycombe', 'Wyre', 'Wyre Forest', 'York']

Sedgemoor Analysis

In [85]:
sedgemoor_df = ukroadaccident[ukroadaccident['District Area'] == 'Sedgemoor']
In [86]:
monthly_sedgemoordata = sedgemoor_df.groupby(['Year', 'Month']).size()

monthly_sedgemoordata = monthly_sedgemoordata.unstack(level=0)

monthly_sedgemoordata.plot(
    marker='o', 
    figsize=(10, 5),
    title='Monthly Road Accidents in Sedgemoor by Year (2019-2022)'
)
plt.xticks(range(0, 12), ['Jan','Feb','Mar','Apr','May','Jun','Jul','Aug','Sep','Oct','Nov','Dec'])
plt.legend(title='Yearly Accident Comparison', bbox_to_anchor = (1.05,1), loc='upper left')
plt.grid(alpha=0.3)
plt.show()
No description has been provided for this image

Insight 1: Compared to the national where we can actually see the clear distribution of accidents per year, in the disctrict of Sedgemoor, the records per year is mostly overlapping with each other and there are incosistent ups and downs.

In [87]:
# Accidents per year
accidents_per_year = sedgemoor_df['Year'].value_counts().sort_index()
accidents_per_month = sedgemoor_df['Month_Number'].value_counts().sort_index()
accidents_per_day = sedgemoor_df['DayofWeek'].value_counts().sort_index()


#PER YEAR 
accidents_per_year.plot(kind='bar')
plt.title('Accidents per Year')
plt.xlabel('Year')
plt.ylabel('Number of Accidents')
plt.show()

month_names = {
    1: 'January', 2: 'February', 3: 'March', 4: 'April',
    5: 'May', 6: 'June', 7: 'July', 8: 'August',
    9: 'September', 10: 'October', 11: 'November', 12: 'December'
}

# Replace numerical months with their names
accidents_per_month.index = accidents_per_month.index.map(month_names)


accidents_per_month.plot(kind='bar')
plt.title('Accidents per Month')
plt.xlabel('Month')
plt.ylabel('Number of Accidents')
plt.show()


#accident per day
day_names = {
    0: 'Monday',
    1: 'Tuesday',
    2: 'Wednesday',
    3: 'Thursday',
    4: 'Friday',
    5: 'Saturday',
    6: 'Sunday'
}

accidents_per_day.index = accidents_per_day.index.map(day_names)
accidents_per_day.plot(kind='bar')
plt.xticks(rotation=45)
print(accidents_per_day)
No description has been provided for this image
No description has been provided for this image 
DayofWeek
Monday       171
Tuesday      181
Wednesday    203
Thursday     193
Friday       205
Saturday     208
Sunday       180
Name: count, dtype: int64
No description has been provided for this image

Insight 2: In terms of Year, the national record is decreasing by year. While in Sedgemoor, year 2019 is the highest, followed by 2021, then 2020, 2022 is the lowest.

Insight 3: In terms of Month, if November is peak month in the national record, in Sedgemoor, September is the peak month. Then both has February as the month with the lowest record.

Insight 4: In terms of Day of the Week, Saturday is the day with highest records in both of them (national record and Wyre) but Friday and Wednesday are also high, then same goes with the lowest which is also both Monday.

Accident Severity in Sedgemoor

In [88]:
severity_sedgemoor  = sedgemoor_df['Accident_Severity'].value_counts()

plt.pie(severity_sedgemoor,
        labels = severity_sedgemoor.index,
        autopct = '%1.1f%%',
        startangle = 90)
plt.gcf().set_size_inches(10,7)
plt.title('Accident Severity Distribution')
plt.show()

severity_2019_sedgemoor = sedgemoor_df[sedgemoor_df['Year'] == 2019]
# severity_2019
lsc_2019 = severity_2019_sedgemoor['Accident_Severity'].value_counts()

plt.pie(lsc_2019,
        labels = lsc_2019.index,
        autopct = '%1.1f%%',
        startangle = 90)
plt.gcf().set_size_inches(10,7)
plt.title('Accident Severity Distribution')
plt.show()
No description has been provided for this image
No description has been provided for this image

Insight 5: Both also has the same distribution of accident severity in the graph, but the record of fatality in Sedgemoor in both charts are slightly higher (above 2%) than the national records.

In [138]:
sedgemoor_df = ukroadaccident[ukroadaccident['District Area'] == 'Sedgemoor']
sedge = list(zip(sedgemoor_df['Latitude'], sedgemoor_df['Longitude']))
m = folium.Map(location=[sedgemoor_df['Latitude'].mean(), sedgemoor_df['Longitude'].mean()], zoom_start=10)

HeatMap(sedge).add_to(m)
# m.save('Sedgemoor_heatmap.html')
m
Out[138]:
Make this Notebook Trusted to load map: File -> Trust Notebook

Stirling Analysis

In [90]:
stirling_df = ukroadaccident[ukroadaccident['District Area'] == 'Stirling']
In [91]:
monthly_stirlingdata = stirling_df.groupby(['Year', 'Month']).size()

monthly_stirlingdata = monthly_stirlingdata.unstack(level=0)

monthly_stirlingdata.plot(
    marker='o', 
    figsize=(10, 5),
    title='Monthly Road Accidents in Stirling by Year (2019-2022)'
)
plt.xticks(range(0, 12), ['Jan','Feb','Mar','Apr','May','Jun','Jul','Aug','Sep','Oct','Nov','Dec'])
plt.legend(title='Yearly Accident Comparison', bbox_to_anchor = (1.05,1), loc='upper left')
plt.grid(alpha=0.3)
plt.show()
No description has been provided for this image

Insight 1: Compared to the national where we can actually see the clear distribution of accidents per year, in Stirling, there is a peak in June 2020. Also, same with Aberdeen City, there is also a consistent zero record in all months on the year 2022 in this district.

In [92]:
# Accidents per year
accidents_per_year = stirling_df['Year'].value_counts().sort_index()
accidents_per_month = stirling_df['Month_Number'].value_counts().sort_index()
accidents_per_day = stirling_df['DayofWeek'].value_counts().sort_index()


#PER YEAR 
accidents_per_year.plot(kind='bar')
plt.title('Accidents per Year')
plt.xlabel('Year')
plt.ylabel('Number of Accidents')
plt.show()

month_names = {
    1: 'January', 2: 'February', 3: 'March', 4: 'April',
    5: 'May', 6: 'June', 7: 'July', 8: 'August',
    9: 'September', 10: 'October', 11: 'November', 12: 'December'
}

# Replace numerical months with their names
accidents_per_month.index = accidents_per_month.index.map(month_names)

accidents_per_month.plot(kind='bar')
plt.title('Accidents per Month')
plt.xlabel('Month')
plt.ylabel('Number of Accidents')
plt.show()


#accident per day
day_names = {
    0: 'Monday',
    1: 'Tuesday',
    2: 'Wednesday',
    3: 'Thursday',
    4: 'Friday',
    5: 'Saturday',
    6: 'Sunday'
}

accidents_per_day.index = accidents_per_day.index.map(day_names)
accidents_per_day.plot(kind='bar')
print(accidents_per_day)
No description has been provided for this image
No description has been provided for this image 
DayofWeek
Monday       119
Tuesday      119
Wednesday    122
Thursday     104
Friday       100
Saturday     129
Sunday       127
Name: count, dtype: int64
No description has been provided for this image

Insight 2: In terms of Year, the national record is decreasing by year. While in Stirling, it is also decreasing per year.

Insight 3: In terms of Month, if November is the peak month in the national record, it is actually the lowest month in Stirling. While the peak month is June.

Insight 4: In terms of Day of the Week, weekends is being the top, but Saturday got the highest record, which is the same with the national analytics. Then the day that has the lowest record in Stirling is Friday, while the national record is Monday, both are weekdays.

Accident Severity in Stirling

In [93]:
severity_stirling  = stirling_df['Accident_Severity'].value_counts()

plt.pie(severity_stirling,
        labels = severity_stirling.index,
        autopct = '%1.1f%%',
        startangle = 90)
plt.gcf().set_size_inches(10,7)
plt.title('Accident Severity Distribution')
plt.show()

severity_2019_stirling = stirling_df[stirling_df['Year'] == 2019]
# severity_2019
lsc_2019 = severity_2019_stirling['Accident_Severity'].value_counts()

plt.pie(lsc_2019,
        labels = lsc_2019.index,
        autopct = '%1.1f%%',
        startangle = 90)
plt.gcf().set_size_inches(10,7)
plt.title('Accident Severity Distribution')
plt.show()
No description has been provided for this image
No description has been provided for this image

Insight 5: Both also has the same distribution of accident severity in the graph, but the rate of fatality and serious injury is slightly higher.

In [141]:
stirling_df = ukroadaccident[ukroadaccident['District Area'] == 'Stirling']
stir = list(zip(stirling_df['Latitude'], stirling_df['Longitude']))
m = folium.Map(location=[stirling_df['Latitude'].mean(), stirling_df['Longitude'].mean()], zoom_start=10)

HeatMap(stir).add_to(m)
# m.save('Stirling_heatmap.html')
m
Out[141]:
Make this Notebook Trusted to load map: File -> Trust Notebook

Rushcliffe Analysis

In [95]:
rushcliffe_df = ukroadaccident[ukroadaccident['District Area'] == 'Rushcliffe']
In [96]:
monthly_rushcliffedata = rushcliffe_df.groupby(['Year', 'Month']).size()

monthly_rushcliffedata = monthly_rushcliffedata.unstack(level=0)

monthly_rushcliffedata.plot(
    marker='o', 
    figsize=(10, 5),
    title='Monthly Road Accidents in Rushcliffe by Year (2019-2022)'
)
plt.xticks(range(0, 12), ['Jan','Feb','Mar','Apr','May','Jun','Jul','Aug','Sep','Oct','Nov','Dec'])
plt.legend(title='Yearly Accident Comparison', bbox_to_anchor = (1.05,1), loc='upper left')
plt.grid(alpha=0.3)
plt.show()
No description has been provided for this image

Insight 1: In the national analytics, it is sequential and decreasing by year. While in Rushcliffe district, the distribution of accidents per month goes up and down by year, and there is a peak in September 2019.

In [97]:
# Accidents per year
accidents_per_year = rushcliffe_df['Year'].value_counts().sort_index()
accidents_per_month = rushcliffe_df['Month_Number'].value_counts().sort_index()
accidents_per_day = rushcliffe_df['DayofWeek'].value_counts().sort_index()


#PER YEAR 
accidents_per_year.plot(kind='bar')
plt.title('Accidents per Year')
plt.xlabel('Year')
plt.ylabel('Number of Accidents')
plt.show()

month_names = {
    1: 'January', 2: 'February', 3: 'March', 4: 'April',
    5: 'May', 6: 'June', 7: 'July', 8: 'August',
    9: 'September', 10: 'October', 11: 'November', 12: 'December'
}

# Replace numerical months with their names
accidents_per_month.index = accidents_per_month.index.map(month_names)


accidents_per_month.plot(kind='bar')
plt.title('Accidents per Month')
plt.xlabel('Month')
plt.ylabel('Number of Accidents')
plt.show()

#accident per day
day_names = {
    0: 'Monday',
    1: 'Tuesday',
    2: 'Wednesday',
    3: 'Thursday',
    4: 'Friday',
    5: 'Saturday',
    6: 'Sunday'
}

accidents_per_day.index = accidents_per_day.index.map(day_names)
accidents_per_day.plot(kind='bar')
print(accidents_per_day)
No description has been provided for this image
No description has been provided for this image 
DayofWeek
Monday       165
Tuesday      195
Wednesday    192
Thursday     185
Friday       174
Saturday     200
Sunday       201
Name: count, dtype: int64
No description has been provided for this image

Insight 2: In terms of Year, both the national record and Rushcliffe analytics is decreasing by year.

Insight 3: In terms of Month, November is the peak month of both analytics, while the lowest point in the National is February and August in Rushcliffe which is very far from each other.

Insight 4: In terms of Day of the Week, Saturday is the peak day in the national, while in Rushcliffe it is both weekends, Sunday is only one record difference from Saturday. Whereas the lowest records of accidents in both analytics is Monday.

Accident Severity in Rushcliffe

In [98]:
severity_rushcliffe  = rushcliffe_df['Accident_Severity'].value_counts()

plt.pie(severity_rushcliffe,
        labels = severity_rushcliffe.index,
        autopct = '%1.1f%%',
        startangle = 90)
plt.gcf().set_size_inches(10,7)
plt.title('Accident Severity Distribution')
plt.show()

severity_2019_rushcliffe = rushcliffe_df[rushcliffe_df['Year'] == 2019]
# severity_2019
lsc_2019 = severity_2019_rushcliffe['Accident_Severity'].value_counts()

plt.pie(lsc_2019,
        labels = lsc_2019.index,
        autopct = '%1.1f%%',
        startangle = 90)
plt.gcf().set_size_inches(10,7)
plt.title('Accident Severity Distribution')
plt.show()
No description has been provided for this image
No description has been provided for this image

Insight 5: Both also has almost the same distribution of accident severity in the graph.

In [142]:
rushcliffe_df = ukroadaccident[ukroadaccident['District Area'] == 'Rushcliffe']
rush = list(zip(rushcliffe_df['Latitude'], rushcliffe_df['Longitude']))
m = folium.Map(location=[rushcliffe_df['Latitude'].mean(), rushcliffe_df['Longitude'].mean()], zoom_start=10)

HeatMap(rush).add_to(m)
# m.save('Rushcliffe_heatmap.html')
m
Out[142]:
Make this Notebook Trusted to load map: File -> Trust Notebook

Dacorum Analysis

In [100]:
dacorum_df = ukroadaccident[ukroadaccident['District Area'] == 'Dacorum']
In [101]:
monthly_dacorumdata = dacorum_df.groupby(['Year', 'Month']).size()

monthly_dacorumdata = monthly_dacorumdata.unstack(level=0)

monthly_dacorumdata.plot(
    marker='o', 
    figsize=(10, 5),
    title='Monthly Road Accidents in Dacorum by Year (2019-2022)'
)
plt.xticks(range(0, 12), ['Jan','Feb','Mar','Apr','May','Jun','Jul','Aug','Sep','Oct','Nov','Dec'])
plt.legend(title='Yearly Accident Comparison', bbox_to_anchor = (1.05,1), loc='upper left')
plt.grid(alpha=0.3)
plt.show()
No description has been provided for this image

Insight 1: In the national analytics, it is sequential and decreasing by year. While in Dacorum, there is a peak in August, 2019, and a lowest point in June 2021.

In [102]:
# Accidents per year
accidents_per_year = dacorum_df['Year'].value_counts().sort_index()
accidents_per_month = dacorum_df['Month_Number'].value_counts().sort_index()
accidents_per_day = dacorum_df['DayofWeek'].value_counts().sort_index()


#PER YEAR 
accidents_per_year.plot(kind='bar')
plt.title('Accidents per Year')
plt.xlabel('Year')
plt.ylabel('Number of Accidents')
plt.show()

month_names = {
    1: 'January', 2: 'February', 3: 'March', 4: 'April',
    5: 'May', 6: 'June', 7: 'July', 8: 'August',
    9: 'September', 10: 'October', 11: 'November', 12: 'December'
}

# Replace numerical months with their names
accidents_per_month.index = accidents_per_month.index.map(month_names)


accidents_per_month.plot(kind='bar')
plt.title('Accidents per Month')
plt.xlabel('Month')
plt.ylabel('Number of Accidents')
plt.show()


#accident per day
day_names = {
    0: 'Monday',
    1: 'Tuesday',
    2: 'Wednesday',
    3: 'Thursday',
    4: 'Friday',
    5: 'Saturday',
    6: 'Sunday'
}

accidents_per_day.index = accidents_per_day.index.map(day_names)
accidents_per_day.plot(kind='bar')
print(accidents_per_day)
No description has been provided for this image
No description has been provided for this image 
DayofWeek
Monday       152
Tuesday      179
Wednesday    212
Thursday     223
Friday       212
Saturday     207
Sunday       214
Name: count, dtype: int64
No description has been provided for this image

Insight 2: In terms of Year, the national record is decreasing per year, while in Dacorum, it is decreasing from 2019-2021, but in 2022. the record of accidents increases slightly.

Insight 3: In terms of Month, November is the peak month of both analytics, while the lowest point in the National is February and August in Dacorum (the same with the analysis in Rushcliffe).

Insight 4: In terms of Day of the Week, Saturday is the peak day in the national, while in Dacorum, it is Thursday. Whereas the lowest records of accidents in both analytics is Monday.

In [103]:
severity_dacorum  = dacorum_df['Accident_Severity'].value_counts()

plt.pie(severity_dacorum,
        labels = severity_dacorum.index,
        autopct = '%1.1f%%',
        startangle = 90)
plt.gcf().set_size_inches(10,7)
plt.title('Accident Severity Distribution')
plt.show()

severity_2019_dacorum = dacorum_df[dacorum_df['Year'] == 2019]
# severity_2019
lsc_2019 = severity_2019_dacorum['Accident_Severity'].value_counts()

plt.pie(lsc_2019,
        labels = lsc_2019.index,
        autopct = '%1.1f%%',
        startangle = 90)
plt.gcf().set_size_inches(10,7)
plt.title('Accident Severity Distribution')
plt.show()
No description has been provided for this image
No description has been provided for this image

Insight 5: Both also has almost the same distribution of accident severity in the graph, but the fatality rate is very much low in Dacorum.

In [144]:
dacorum_df = ukroadaccident[ukroadaccident['District Area'] == 'Dacorum']
dac = list(zip(dacorum_df['Latitude'], dacorum_df['Longitude']))
m = folium.Map(location=[dacorum_df['Latitude'].mean(), dacorum_df['Longitude'].mean()], zoom_start=10)

HeatMap(dac).add_to(m)
# m.save('Dacorum_heatmap.html')
m
Out[144]:
Make this Notebook Trusted to load map: File -> Trust Notebook
In [143]:
# pd.DataFrame(ukroadaccident['District Area'].unique(), columns=['District Area'])
# ukroadaccident['District Area'].value_counts()

for area in ukroadaccident['District Area'].unique():
    print(area)

Kensington and Chelsea
Westminster
Richmond upon Thames
Hammersmith and Fulham
Hounslow
Tower Hamlets
City of London
Southwark
Camden
Hackney
Islington
Barnet
Brent
Haringey
Merton
Ealing
Enfield
Greenwich
Newham
London Airport (Heathrow)
Hillingdon
Waltham Forest
Redbridge
Barking and Dagenham
Bromley
Havering
Croydon
Lambeth
Wandsworth
Sutton
Bexley
Lewisham
Harrow
Kingston upon Thames
Barrow-in-Furness
South Lakeland
Carlisle
Eden
Allerdale
Copeland
Fylde
Blackpool
Wyre
Lancaster
Chorley
West Lancashire
South Ribble
Preston
Blackburn with Darwen
Hyndburn
Ribble Valley
Burnley
Pendle
Rossendale
Wirral
Liverpool
Sefton
Knowsley
St. Helens
Manchester
Salford
Tameside
Stockport
Bolton
Wigan
Trafford
Bury
Rochdale
Oldham
Chester
Vale Royal
Halton
Warrington
Crewe and ntwich
Ellesmere Port and Neston
Congleton
Macclesfield
Crewe and Nantwich
Blyth Valley
Sunderland
Gateshead
North Tyneside
Tynedale
Berwick-upon-Tweed
Newcastle upon Tyne
Castle Morpeth
South Tyneside
Wansbeck
Alnwick
Wear Valley
Sedgefield
Durham
Easington
Chester-le-Street
Derwentside
Teesdale
Darlington
Craven
Hambleton
Richmondshire
York
Harrogate
Selby
Scarborough
Ryedale
Kirklees
Leeds
Wakefield
Bradford
Calderdale
Doncaster
Barnsley
Sheffield
Rotherham
North East Lincolnshire
North Lincolnshire
East Riding of Yorkshire
Kingston upon Hull, City of
Redcar and Cleveland
Hartlepool
Middlesbrough
Stockton-on-Tees
Birmingham
Wolverhampton
Walsall
Dudley
Sandwell
Solihull
Coventry
Newcastle-under-Lyme
Stafford
East Staffordshire
South Staffordshire
Lichfield
Stoke-on-Trent
Cannock Chase
Tamworth
Staffordshire Moorlands
Malvern Hills
Worcester
Wychavon
Wyre Forest
Redditch
Bromsgrove
Herefordshire, County of
Telford and Wrekin
South Shropshire
Bridgnorth
Shrewsbury and Atcham
North Shropshire
Oswestry
Warwick
Stratford-upon-Avon
Rugby
North Warwickshire
Nuneaton and Bedworth
Amber Valley
Erewash
Bolsover
High Peak
Derbyshire Dales
North East Derbyshire
Chesterfield
Derby
South Derbyshire
Ashfield
Mansfield
Newark and Sherwood
Bassetlaw
Gedling
Rushcliffe
Nottingham
Broxtowe
South Holland
South Kesteven
North Kesteven
East Lindsey
Boston
West Lindsey
Lincoln
Hinckley and Bosworth
Harborough
Leicester
Rutland
Charnwood
Melton
North West Leicestershire
Blaby
Oadby and Wigston
South Northamptonshire
Northampton
Kettering
Corby
Daventry
East Northamptonshire
Wellingborough
South Cambridgeshire
Fenland
Huntingdonshire
Peterborough
East Cambridgeshire
Cambridge
King's Lynn and West Norfolk
Norwich
Breckland
South Norfolk
Broadland
Great Yarmouth
North Norfolk
St. Edmundsbury
Babergh
Mid Suffolk
Forest Heath
Suffolk Coastal
Ipswich
Waveney
Luton
Mid Bedfordshire
Bedford
South Bedfordshire
East Hertfordshire
North Hertfordshire
Welwyn Hatfield
St. Albans
Hertsmere
Three Rivers
Watford
Dacorum
Broxbourne
Stevege
Stevenage
Epping Forest
Brentwood
Castle Point
Uttlesford
Chelmsford
Basildon
Tendring
Colchester
Thurrock
Braintree
Rochford
Maldon
Southend-on-Sea
Harlow
Aylesbury Vale
Wycombe
South Bucks
Chiltern
Slough
Windsor and Maidenhead
South Oxfordshire
Milton Keynes
Cherwell
Vale of White Horse
West Oxfordshire
Oxford
Reading
West Berkshire
Wokingham
Bracknell Forest
Eastleigh
Test Valley
Portsmouth
Winchester
Basingstoke and Deane
Southampton
Fareham
Hart
Gosport
Havant
Isle of Wight
East Hampshire
Rushmoor
New Forest
Elmbridge
Mole Valley
Guildford
Spelthorne
Epsom and Ewell
Runnymede
Woking
Reigate and Banstead
Surrey Heath
Waverley
Tandridge
Medway
Thanet
Shepway
Ashford
Maidstone
Dover
Tunbridge Wells
Canterbury
Tonbridge and Malling
Swale
Dartford
Sevenoaks
Gravesham
Brighton and Hove
Chichester
Rother
Lewes
Wealden
Eastbourne
Hastings
Crawley
Horsham
Mid Sussex
Arun
Adur
Worthing
Kerrier
Restormel
Carrick
Penwith
North Cornwall
West Devon
Torridge
Caradon
North Devon
Mid Devon
Exeter
East Devon
Teignbridge
Plymouth
South Hams
Torbay
Bristol, City of
South Somerset
Sedgemoor
Mendip
Taunton Deane
South Gloucestershire
North Somerset
Bath and North East Somerset
West Somerset
Stroud
Tewkesbury
Gloucester
Cotswold
Cheltenham
Forest of Dean
Salisbury
Kennet
West Wiltshire
North Wiltshire
Swindon
Bournemouth
Poole
East Dorset
West Dorset
Weymouth and Portland
Christchurch
Purbeck
North Dorset
Isle of Anglesey
Conwy
Denbighshire
Flintshire
Wrexham
Gwynedd
Monmouthshire
Newport
Caerphilly
Torfaen
Blaeu Gwent
Blaenau Gwent
Merthyr Tydfil
The Vale of Glamorgan
Cardiff
Rhondda, Cynon, Taff
Swansea
Bridgend
Neath Port Talbot
Carmarthenshire
Ceredigion
Pembrokeshire
Powys
Highland
Western Isles
Orkney Islands
Shetland Islands
Aberdeenshire
Moray
Aberdeen City
Dundee City
Angus
Perth and Kinross
Fife
East Lothian
Midlothian
Scottish Borders
Edinburgh, City of
West Lothian
Falkirk
Clackmannanshire
Stirling
Clackmannshire
Glasgow City
East Dunbartonshire
North Larkshire
North Lanarkshire
Renfrewshire
East Renfrewshire
East Ayrshire
South Larkshire
Inverclyde
West Dunbartonshire
Argyll and Bute
South Lanarkshire
North Ayrshire
South Ayrshire
Dumfries and Galloway
Cheshire East
Cheshire West and Chester
Northumberland
County Durham
Shropshire
Central Bedfordshire
Cornwall
Wiltshire

Dundee City Analysis

In [106]:
dundee_df = ukroadaccident[ukroadaccident['District Area'] == 'Dundee City']
In [107]:
monthly_dundeedata = dundee_df.groupby(['Year', 'Month']).size()

monthly_dundeedata = monthly_dundeedata.unstack(level=0)

monthly_dundeedata.plot(
    marker='o', 
    figsize=(10, 5),
    title='Monthly Road Accidents in Dundee City by Year (2019-2022)'
)
plt.xticks(range(0, 12), ['Jan','Feb','Mar','Apr','May','Jun','Jul','Aug','Sep','Oct','Nov','Dec'])
plt.legend(title='Yearly Accident Comparison', bbox_to_anchor = (1.05,1), loc='upper left')
plt.grid(alpha=0.3)
plt.show()
No description has been provided for this image

Insight 1: In the national analytics, it is sequential and decreasing by year. While in Dundee City, the record is quite inconsistent in 2019,2020, and 2021. Also, same with Aberdeen City, and Stirling, it doesnt have any records in 2022.

In [108]:
# Accidents per year
accidents_per_year = dundee_df['Year'].value_counts().sort_index()
accidents_per_month = dundee_df['Month_Number'].value_counts().sort_index()
accidents_per_day = dundee_df['DayofWeek'].value_counts().sort_index()


#PER YEAR 
accidents_per_year.plot(kind='bar')
plt.title('Accidents per Year')
plt.xlabel('Year')
plt.ylabel('Number of Accidents')
plt.show()

month_names = {
    1: 'January', 2: 'February', 3: 'March', 4: 'April',
    5: 'May', 6: 'June', 7: 'July', 8: 'August',
    9: 'September', 10: 'October', 11: 'November', 12: 'December'
}

# Replace numerical months with their names
accidents_per_month.index = accidents_per_month.index.map(month_names)


accidents_per_month.plot(kind='bar')
plt.title('Accidents per Month')
plt.xlabel('Month')
plt.ylabel('Number of Accidents')
plt.show()


#accident per day
day_names = {
    0: 'Monday',
    1: 'Tuesday',
    2: 'Wednesday',
    3: 'Thursday',
    4: 'Friday',
    5: 'Saturday',
    6: 'Sunday'
}

accidents_per_day.index = accidents_per_day.index.map(day_names)
accidents_per_day.plot(kind='bar')
print(accidents_per_day)
No description has been provided for this image
No description has been provided for this image 
DayofWeek
Monday        81
Tuesday      117
Wednesday    133
Thursday     116
Friday       109
Saturday     143
Sunday       100
Name: count, dtype: int64
No description has been provided for this image

Insight 2: In terms of Year, if the national record is decreasing per year, the records in Dundee City is increasing per year (2019-2021), then it is a sudden drop with literally zero record in 2022.

Insight 3: In terms of Month, both analysis has November as the peak month, and February as the lowest month.

Insight 4: In terms of Day of the Week, both has Saturday as the peak day, and Monday as the lowest.

Accident Severity in Dundee City

In [109]:
severity_dundee  = dundee_df['Accident_Severity'].value_counts()

plt.pie(severity_dundee,
        labels = severity_dundee.index,
        autopct = '%1.1f%%',
        startangle = 90)
plt.gcf().set_size_inches(10,7)
plt.title('Accident Severity Distribution')
plt.show()

severity_2019_dundee = dundee_df[dundee_df['Year'] == 2019]
# severity_2019
lsc_2019 = severity_2019_dundee['Accident_Severity'].value_counts()

plt.pie(lsc_2019,
        labels = lsc_2019.index,
        autopct = '%1.1f%%',
        startangle = 90)
plt.gcf().set_size_inches(10,7)
plt.title('Accident Severity Distribution')
plt.show()
No description has been provided for this image
No description has been provided for this image

Insight 5: Both also has almost the same distribution of accident severity in the graph, but the serious injury is very much higher in Dundee City.

In [145]:
dundee_df = ukroadaccident[ukroadaccident['District Area'] == 'Dundee City']
dun = list(zip(dundee_df['Latitude'], dundee_df['Longitude']))
m = folium.Map(location=[dundee_df['Latitude'].mean(), dundee_df['Longitude'].mean()], zoom_start=10)

HeatMap(dun).add_to(m)
# m.save('Dundee_heatmap.html')
m
Out[145]:
Make this Notebook Trusted to load map: File -> Trust Notebook

Glasgow City Analysis

In [111]:
glasgow_df = ukroadaccident[ukroadaccident['District Area'] == 'Glasgow City']
In [112]:
monthly_glasgowdata = glasgow_df.groupby(['Year', 'Month']).size()

monthly_glasgowdata = monthly_glasgowdata.unstack(level=0)

monthly_glasgowdata.plot(
    marker='o', 
    figsize=(10, 5),
    title='Monthly Road Accidents in Glasgow City by Year (2019-2022)'
)
plt.xticks(range(0, 12), ['Jan','Feb','Mar','Apr','May','Jun','Jul','Aug','Sep','Oct','Nov','Dec'])
plt.legend(title='Yearly Accident Comparison', bbox_to_anchor = (1.05,1), loc='upper left')
plt.grid(alpha=0.3)
plt.show()
No description has been provided for this image

Insight 1: In the national analytics, it is sequential and decreasing by year. While in Glasgow City, the record is quite inconsistent in 2019,2020, and 2021. Also, same with Aberdeen City, Stirling, and Dundee City, this district has zero record in 2022.

In [113]:
# Accidents per year
accidents_per_year = glasgow_df['Year'].value_counts().sort_index()
accidents_per_month = glasgow_df['Month_Number'].value_counts().sort_index()
accidents_per_day = glasgow_df['DayofWeek'].value_counts().sort_index()


#PER YEAR 
accidents_per_year.plot(kind='bar')
plt.title('Accidents per Year')
plt.xlabel('Year')
plt.ylabel('Number of Accidents')
plt.show()

month_names = {
    1: 'January', 2: 'February', 3: 'March', 4: 'April',
    5: 'May', 6: 'June', 7: 'July', 8: 'August',
    9: 'September', 10: 'October', 11: 'November', 12: 'December'
}

# Replace numerical months with their names
accidents_per_month.index = accidents_per_month.index.map(month_names)


accidents_per_month.plot(kind='bar')
plt.title('Accidents per Month')
plt.xlabel('Month')
plt.ylabel('Number of Accidents')
plt.show()


#accident per day
day_names = {
    0: 'Monday',
    1: 'Tuesday',
    2: 'Wednesday',
    3: 'Thursday',
    4: 'Friday',
    5: 'Saturday',
    6: 'Sunday'
}

accidents_per_day.index = accidents_per_day.index.map(day_names)
accidents_per_day.plot(kind='bar')
print(accidents_per_day)
No description has been provided for this image
No description has been provided for this image 
DayofWeek
Monday       522
Tuesday      732
Wednesday    709
Thursday     735
Friday       738
Saturday     820
Sunday       686
Name: count, dtype: int64
No description has been provided for this image

Insight 2: In terms of Year, both analysis record of accidents decreases per year, but Glasgow City has a sudden drop to zero in 2022.

Insight 3: In terms of Month, both analysis has November as the peak month, while the lowest record in the national is February and Glasgow City has June.

Insight 4: In terms of Day of the Week, both has Saturday as the peak day, and Monday as the lowest.

Accident Severity in Glasgow City

In [114]:
severity_glasgow  = glasgow_df['Accident_Severity'].value_counts()

plt.pie(severity_glasgow,
        labels = severity_glasgow.index,
        autopct = '%1.1f%%',
        startangle = 90)
plt.gcf().set_size_inches(10,7)
plt.title('Accident Severity Distribution')
plt.show()

severity_2019_glasgow = glasgow_df[glasgow_df['Year'] == 2019]
# severity_2019
lsc_2019 = severity_2019_glasgow['Accident_Severity'].value_counts()

plt.pie(lsc_2019,
        labels = lsc_2019.index,
        autopct = '%1.1f%%',
        startangle = 90)
plt.gcf().set_size_inches(10,7)
plt.title('Accident Severity Distribution')
plt.show()
No description has been provided for this image
No description has been provided for this image

Insight 5: Both also has almost the same distribution of accident severity in the graph.

In [146]:
glasgow_df = ukroadaccident[ukroadaccident['District Area'] == 'Glasgow City']
glas = list(zip(glasgow_df['Latitude'], glasgow_df['Longitude']))
m = folium.Map(location=[glasgow_df['Latitude'].mean(), glasgow_df['Longitude'].mean()], zoom_start=10)

HeatMap(glas).add_to(m)
# m.save('Glasgow_heatmap.html')
m
Out[146]:
Make this Notebook Trusted to load map: File -> Trust Notebook

Barking and Dagenham Analysis

In [116]:
barda_df = ukroadaccident[ukroadaccident['District Area'] == 'Barking and Dagenham']
In [117]:
monthly_bardadata = barda_df.groupby(['Year', 'Month']).size()

monthly_bardadata = monthly_bardadata.unstack(level=0)

monthly_bardadata.plot(
    marker='o', 
    figsize=(10, 5),
    title='Monthly Road Accidents in Barking and Dagenham by Year (2019-2022)'
)
plt.xticks(range(0, 12), ['Jan','Feb','Mar','Apr','May','Jun','Jul','Aug','Sep','Oct','Nov','Dec'])
plt.legend(title='Yearly Accident Comparison', bbox_to_anchor = (1.05,1), loc='upper left')
plt.grid(alpha=0.3)
plt.show()
No description has been provided for this image

Insight 1: In the national analytics, it is sequential and decreasing by year. While in Barking and Dagenham, there is a lowest point in March 2021, and a peak in September 2020.

In [118]:
# Accidents per year
accidents_per_year = barda_df['Year'].value_counts().sort_index()
accidents_per_month = barda_df['Month_Number'].value_counts().sort_index()
accidents_per_day = barda_df['DayofWeek'].value_counts().sort_index()


#PER YEAR 
accidents_per_year.plot(kind='bar')
plt.title('Accidents per Year')
plt.xlabel('Year')
plt.ylabel('Number of Accidents')
plt.show()

month_names = {
    1: 'January', 2: 'February', 3: 'March', 4: 'April',
    5: 'May', 6: 'June', 7: 'July', 8: 'August',
    9: 'September', 10: 'October', 11: 'November', 12: 'December'
}

# Replace numerical months with their names
accidents_per_month.index = accidents_per_month.index.map(month_names)


accidents_per_month.plot(kind='bar')
plt.title('Accidents per Month')
plt.xlabel('Month')
plt.ylabel('Number of Accidents')
plt.show()


#accident per day
day_names = {
    0: 'Monday',
    1: 'Tuesday',
    2: 'Wednesday',
    3: 'Thursday',
    4: 'Friday',
    5: 'Saturday',
    6: 'Sunday'
}

accidents_per_day.index = accidents_per_day.index.map(day_names)
accidents_per_day.plot(kind='bar')
print(accidents_per_day)
No description has been provided for this image
No description has been provided for this image 
DayofWeek
Monday       198
Tuesday      256
Wednesday    266
Thursday     254
Friday       239
Saturday     306
Sunday       250
Name: count, dtype: int64
No description has been provided for this image

Insight 2: In terms of Year, the national record decreases per year, while in Barking and Dagenham, it looks like it has a pattern of decreasing-increasing-decreasing-increasing.

Insight 3: In terms of Month, both analysis has November as the peak month, while the lowest record is February in both as well.

Insight 4: In terms of Day of the Week, both has Saturday as the peak day, and Monday as the lowest.

Accident Severity in Barking and Dagenham

In [119]:
severity_barda  = barda_df['Accident_Severity'].value_counts()

plt.pie(severity_barda,
        labels = severity_barda.index,
        autopct = '%1.1f%%',
        startangle = 90)
plt.gcf().set_size_inches(10,7)
plt.title('Accident Severity Distribution')
plt.show()

severity_2019_barda = barda_df[barda_df['Year'] == 2019]
# severity_2019
lsc_2019 = severity_2019_barda['Accident_Severity'].value_counts()

plt.pie(lsc_2019,
        labels = lsc_2019.index,
        autopct = '%1.1f%%',
        startangle = 90)
plt.gcf().set_size_inches(10,7)
plt.title('Accident Severity Distribution')
plt.show()
No description has been provided for this image
No description has been provided for this image

Insight 5: Both also has almost the same distribution of accident severity in the graph, but the serious injury is much lower in Barking and Dagenham district.

In [147]:
barda_df = ukroadaccident[ukroadaccident['District Area'] == 'Barking and Dagenham']
bark = list(zip(barda_df['Latitude'], barda_df['Longitude']))
m = folium.Map(location=[barda_df['Latitude'].mean(), barda_df['Longitude'].mean()], zoom_start=10)

HeatMap(bark).add_to(m)
# m.save('BarkingAndDagenham_heatmap.html')
m
Out[147]:
Make this Notebook Trusted to load map: File -> Trust Notebook

Kingston upon Thames Analysis

In [121]:
kingtha_df = ukroadaccident[ukroadaccident['District Area'] == 'Kingston upon Thames']
In [122]:
monthly_kingthadata = kingtha_df.groupby(['Year', 'Month']).size()

monthly_kingthadata = monthly_kingthadata.unstack(level=0)

monthly_kingthadata.plot(
    marker='o', 
    figsize=(10, 5),
    title='Monthly Road Accidents in Kingston upon Thames by Year (2019-2022)'
)
plt.xticks(range(0, 12), ['Jan','Feb','Mar','Apr','May','Jun','Jul','Aug','Sep','Oct','Nov','Dec'])
plt.legend(title='Yearly Accident Comparison', bbox_to_anchor = (1.05,1), loc='upper left')
plt.grid(alpha=0.3)
plt.show()
No description has been provided for this image

Insight 1: In the national analytics, it is sequential and decreasing by year. While in Kingston upon Thames, there is a peak in September 2021, and a lowest point in April 2021.

In [123]:
# Accidents per year
accidents_per_year = kingtha_df['Year'].value_counts().sort_index()
accidents_per_month = kingtha_df['Month_Number'].value_counts().sort_index()
accidents_per_day = kingtha_df['DayofWeek'].value_counts().sort_index()


#PER YEAR 
accidents_per_year.plot(kind='bar')
plt.title('Accidents per Year')
plt.xlabel('Year')
plt.ylabel('Number of Accidents')
plt.show()

month_names = {
    1: 'January', 2: 'February', 3: 'March', 4: 'April',
    5: 'May', 6: 'June', 7: 'July', 8: 'August',
    9: 'September', 10: 'October', 11: 'November', 12: 'December'
}

# Replace numerical months with their names
accidents_per_month.index = accidents_per_month.index.map(month_names)


accidents_per_month.plot(kind='bar')
plt.title('Accidents per Month')
plt.xlabel('Month')
plt.ylabel('Number of Accidents')
plt.show()


#accident per day
day_names = {
    0: 'Monday',
    1: 'Tuesday',
    2: 'Wednesday',
    3: 'Thursday',
    4: 'Friday',
    5: 'Saturday',
    6: 'Sunday'
}

accidents_per_day.index = accidents_per_day.index.map(day_names)
accidents_per_day.plot(kind='bar')
print(accidents_per_day)
No description has been provided for this image
No description has been provided for this image 
DayofWeek
Monday       130
Tuesday      219
Wednesday    209
Thursday     229
Friday       230
Saturday     209
Sunday       162
Name: count, dtype: int64
No description has been provided for this image

Insight 2: In terms of Year, the national record decreases per year, while the record in Kingston upon Thames increases by year in 3 years only because it decreases by 2022.

Insight 3: In terms of Month, both analysis has November as the peak month, while the lowest record in the national is February and in Kingston upon Thames is January.

Insight 4: In terms of Day of the Week, both has Monday as the lowest. The highest day in the national is Saturday, while in Kingston upon Thames, Friday is only one record higher from Thursday.

Accident Severity in Kingston upon Thames

In [124]:
severity_kingtha  = kingtha_df['Accident_Severity'].value_counts()

plt.pie(severity_kingtha,
        labels = severity_kingtha.index,
        autopct = '%1.1f%%',
        startangle = 90)
plt.gcf().set_size_inches(10,7)
plt.title('Accident Severity Distribution')
plt.show()

severity_2019_kingtha = kingtha_df[kingtha_df['Year'] == 2019]
# severity_2019
lsc_2019 = severity_2019_kingtha['Accident_Severity'].value_counts()

plt.pie(lsc_2019,
        labels = lsc_2019.index,
        autopct = '%1.1f%%',
        startangle = 90)
plt.gcf().set_size_inches(10,7)
plt.title('Accident Severity Distribution')
plt.show()
No description has been provided for this image
No description has been provided for this image

Insight 5: Both also has almost the same distribution of accident severity in the graph, but the overall analysis in Kingston upon Thames is much lower.

In [148]:
kingtha_df = ukroadaccident[ukroadaccident['District Area'] == 'Kingston upon Thames']
thame = list(zip(kingtha_df['Latitude'], kingtha_df['Longitude']))
m = folium.Map(location=[kingtha_df['Latitude'].mean(), kingtha_df['Longitude'].mean()], zoom_start=10)

HeatMap(thame).add_to(m)
# m.save('KingstonUponThames_heatmap.html')
m
Out[148]:
Make this Notebook Trusted to load map: File -> Trust Notebook
In [126]:
# di po nagana sa akin yung heatmap po dito sa jupyter, ayaw po magshow :(
# gumana na po hehehehhe

Data Visualization with Seaborn

In [127]:
import seaborn as sns
In [128]:
ukroadaccident.dtypes
Out[128]:
Index                            category
Accident_Severity                category
Accident Date              datetime64[ns]
Latitude                          float64
Light_Conditions                 category
District Area                    category
Longitude                         float64
Number_of_Casualties                int64
Number_of_Vehicles                  int64
Road_Surface_Conditions          category
Road_Type                        category
Urban_or_Rural_Area              category
Weather_Conditions               category
Vehicle_Type                     category
Year                             category
Month_Number                        int32
Month                            category
Day                                 int32
DayofWeek                           int32
dtype: object
In [129]:
# group1 = ukroadaccident.groupby('Urban_or_Rural_Area').value_counts()
# group2 = ukroadaccident.groupby('Vehicle_Type').value_counts()
# group = sns.barplot(x='group1', y='group2', data=ukroadaccident)
# plt.show()
In [130]:
group = sns.barplot(x='DayofWeek', y='Number_of_Casualties', data=ukroadaccident)
plt.show()
No description has been provided for this image
In [131]:
group = sns.boxplot(x='DayofWeek', y='Number_of_Casualties', data=ukroadaccident)
plt.show()
No description has been provided for this image
In [132]:
group = sns.histplot(x='DayofWeek', y='Number_of_Vehicles', data=ukroadaccident)
plt.show()
No description has been provided for this image
In [133]:
group = sns.histplot(x='Accident_Severity', y='Urban_or_Rural_Area', data=ukroadaccident)
plt.show()
No description has been provided for this image
In [ ]: